def save(DATASET, NUM_FOLD, ENSEMBLE, ALGORITHM, sparse=False):
NUM_GENES = None
SEED = 1313
NTREES = 300
SUB_CROSS = 3
METRIC = 'NDCG'
if DATASET == '2003_td_dataset':
NUM_GENES = 64
elif DATASET == 'web10k':
NUM_GENES = 136
elif DATASET == 'yahoo':
NUM_GENES = 700
elif DATASET in ['movielens', 'lastfm', 'bibsonomy', 'youtube']:
NUM_GENES = 13
else:
print('DATASET INVÁLIDO')
X_train, y_train, query_id_train = l2rCodesSerial.load_L2R_file(
'./dataset/' + DATASET + '/Fold' + NUM_FOLD + '/Norm.' + 'train' +
'.txt', '1' * NUM_GENES, sparse)
# X_test, y_test, query_id_test = l2rCodesSerial.load_L2R_file(
# './dataset/' + DATASET + '/Fold' + NUM_FOLD + '/Norm.' + 'test' + '.txt', '1' * NUM_GENES, sparse)
scoreTest = [0] * len(y_train)
model = l2rCodesSerial.getTheModel(1, NTREES, 0.3, SEED, DATASET)
model.fit(X_train, y_train)
resScore = model.predict(X_train)
c = 0
for i in resScore:
scoreTest[c] = i
c = c + 1
ndcg, queries = l2rCodesSerial.getEvaluation(scoreTest, query_id_train,
y_train, DATASET, METRIC,
"test")
f = open(
'./baselines/' + DATASET + '/Fold' + NUM_FOLD + '/' + ALGORITHM +
'train.txt', "w+")
for i in range(len(queries)):
f.write(str(queries[i]) + '\n')
# f.write(str(queries[i]))
f.close()
def getPrecisionAndQueries(individuo, NUM_GENES, X_train, y_train, X_test,
y_test, query_id_train, ENSEMBLE, NTREES, SEED,
DATASET, METRIC):
# list_mask = list(individuo)
# features = []
# for i in range(NUM_GENES):
# if list_mask[i] == '1':
# features.append(i)
# X_train_ind = X_train[:, features]
# # X_test_ind = X_test[:, features]
#
# queriesList = l2rCodesSerial.getQueries(query_id_train)
# scoreTrain = [0] * len(y_train)
#
# kf = model_selection.KFold(n_splits=3, shuffle=True, random_state=SEED)
# XF_train_index = []
# XF_test_index = []
#
# for qtrain_index, qtest_index in kf.split(queriesList):
#
# del XF_train_index[:]
# del XF_test_index[:]
#
# for qtr in qtrain_index:
# XF_train_index = XF_train_index + queriesList[qtr]
# for qts in qtest_index:
# XF_test_index = XF_test_index + queriesList[qts]
#
# XF_train, XF_test = X_train_ind[XF_train_index], X_train_ind[XF_test_index] ####
# yf_train, yf_test = y_train[XF_train_index], y_train[XF_test_index]
# # qf_train, qf_test = query_id_train[XF_train_index], query_id_train[XF_test_index]
#
# # resScore = None
# model = l2rCodesSerial.getTheModel(ENSEMBLE, NTREES, 0.3, SEED, DATASET)
#
# model.fit(XF_train, yf_train)
# resScore = model.predict(XF_test)
#
# # sortRespRel=[rel for (score, rel) in sorted(zip(resScore,yf_test), key=lambda pair: pair[0], reverse=True)]
# c = 0
# for i in XF_test_index:
# scoreTrain[i] = resScore[c]
# c = c + 1
# ndcg, queries = l2rCodesSerial.getEvaluation(scoreTrain, query_id_train, y_train, DATASET, METRIC, "test")
# return ndcg, queries, scoreTrain
list_mask = list(individuo)
features = []
for i in range(NUM_GENES):
if list_mask[i] == '1':
features.append(i)
X_train_ind = X_train[:, features]
scoreTrain = [0] * len(y_train)
# model = linear_model.LinearRegression(n_jobs=-1)
model = l2rCodesSerial.getTheModel(ENSEMBLE, NTREES, 0.3, SEED, DATASET)
model.fit(X_train_ind, y_train)
resScore = model.predict(X_train_ind)
# sortRespRel=[rel for (score, rel) in sorted(zip(resScore,yf_test), key=lambda pair: pair[0], reverse=True)]
c = 0
for i in range(len(resScore)):
scoreTrain[i] = resScore[c]
c = c + 1
ndcg, queries = l2rCodesSerial.getEvaluation(scoreTrain, query_id_train,
y_train, DATASET, METRIC,
"test")
return ndcg, queries, scoreTrain
# NSGA best IND RF
ENSEMBLE = 1 # random forest
NTREES = 20
SEED = 1887
NUM_FOLD = '0'
METRIC = "NDCG"
sparse = True
ALGORITHM = 'rf'
print('reading and training NSGA bestind')
X_train, y_train, query_id_train = l2rCodesSerial.load_L2R_file(
'./dataset/' + dataset + '/' + NUM_FOLD + '.' + 'train', bestindNSGA, sparse)
X_test, y_test, query_id_test = l2rCodesSerial.load_L2R_file(
'./dataset/' + dataset + '/' + NUM_FOLD + '.' + 'test', bestindNSGA, sparse)
# print(len(X_train[0]))
model = l2rCodesSerial.getTheModel(ENSEMBLE, NTREES, 0.3, SEED, dataset)
model.fit(X_train, y_train)
resScore = model.predict(X_test)
scoreTest = [0] * len(y_test)
c = 0
for i in resScore:
scoreTest[c] = i
c = c + 1
ndcg, queries = l2rCodesSerial.getEvaluation(scoreTest, query_id_test, y_test, dataset, METRIC, "test")
ndcgs.append(queries)
diversitys.append(evaluateIndividuoSerial.getDiversity(scoreTest, y_test, query_id_test))
noveltys.append(evaluateIndividuoSerial.getNovelty(scoreTest, y_test, query_id_test))
# SPEA best IND RF